amit-bendor talks with dana-maman, a non-developer who builds apps with Claude, about her “second brain”. She rebuilt her notes on andrej-karpathy’s LLM Wiki pattern (raw / wiki / schema), then grew it into a personal “operating system” holding meetings, projects, apps, design system, and writing. The core lessons are about trust: instructions aren’t enough, so enforce rules with hooks and loops, verify facts and links by reading sources, and investigate every failure. They also cover file structure, knowledge rot, retrieval beyond ~100 sources (multiple query guesses, ranking candidates, RAG/hybrid), and Google’s OKF format. Getting started: let the AI interview you, sketch your life domains on paper, start simple and restructure freely.
Show: osim-tochna
Concepts: llm-wiki · second-brain · context-engineering · wiki-retrieval · ai-guardrails · ai-verification · knowledge-rot · open-knowledge-format